Using watermarking solution on the un-compressed domain is elegant as the watermarking is done standalone in the client device, and as done in the un-compressed domain, it generally offers the best watermark insertion's capability and better invisibility.
But working after the decoder (for un-compressed watermarking) offers no interaction with the security elements which are linked to the descrambling of the data, thus occurring before the decoder.
The document U.S. 2010/128871 describes a solution in which a secondary stream is generated comprising the data allowing to reconstruct the main stream and at the same time to marl the reconstructed stream. As a consequence, the same stream contains the scrambling data and the watermarking data. At the reception, this stream is processed as one set of data to be replaced in the modified stream.
The document EP 2 391 125 describes a solution to allow an individual marking (at the reception device) based on a stream common to all devices. The control object contains the original value, an alternate value and a location. The security unit determines a mathematical operation to be applied on the alternate value to retrieve the original value. The mathematical operation is changed according to an internal parameter of the reception device so that the mathematical operation will be unique per device, allowing to track this device if the stream of descrambling data is analyzed.